PPP accuracy improves with the length of the data collection period. A minimum period of good quality GPS data (no loss-of-lock) is required to permit convergence and/or resolving ambiguities which in turn can improve the accuracy of the entire dataset. The minimum period and the accuracy attainable will depend on the type of GPS equipment, the site (multipath, obstructions) and atmospheric conditions.
Extending the data collection past this minimum period should further improve accuracy, but more so with dual-frequency receivers than with single frequency.
The duration of data collection should be decided according to the accuracy required. Users should test their equipment for PPP at a known reference point. Here are the results of some in-house analysis and testing.
Analysis of static PPP processing results of 186 24-hr datasets from the high-end dual-frequency receivers setup on our (NRCan) reference stations show that 4 cm accuracy was attained within 2 hours and accuracy continued to improve until it leveled off at the 1 cm level after 12 hours).
Analysis of 554 user-submitted 24-hr dual-frequency datasets show a similar graph however a longer period was required (24 hours) before accuracy truly leveled off.
Because we do not know the true coordinates for the user-supplied datasets but can assume that the accuracy attained then would be close to centimeter level possibly sub-centimeter.
Note: The quicker leveling off in the NRCan graph could be attributed to better quality GPS equipment and better sites.
Limited in-house testing showed one of our mapping-grade single-frequency receiver capable of repeatedly attaining 20 cm accuracy (horizontal and vertical) with 2 hours of data under ideal conditions. Extending past the 2 hours did not appear to improve on this 20 cm threshold.
The use of PPP is not recommended for recreational receivers, they are generally of much lower quality. Our $500 recreational receiver gave disappointing results requiring anywhere from 3 hours to 6 hours to attain 1 m accuracy (under ideal conditions).
As you can see results can vary, testing your equipment is the only way to know for sure.